AI Outside Tech: A Look At The Returns And The Rising Costs

🔍 Read the full analysis: AI Outside Tech: A Look At The Returns And The Rising Costs on ThorstenMeyerAI.com

Buying for a business?Offer from Amazon

Get business pricing on monitors, keyboards and dev gear

• Business-only prices and quantity discounts
• Tax-exempt purchasing
• Multiple users, one account, clear invoices

Create a free Business account →Free to register

As an affiliate, we earn on qualifying purchases.Hide

Apollo chief economist Torsten Slok said an October 24 note found no visible financial payoff from AI outside the technology sector, as corporate data show flat revenue per worker and margins for most S&P 500 companies while AI infrastructure bills grow. The finding describes aggregate trends, not a measured causal effect of AI, and comes as Meta reports roughly $825 billion in future commitments, largely tied to data centers, servers and cloud capacity.

Apollo compared the seven companies it groups as the Magnificent 7 with the other 493 S&P 500 firms. Its charts show inflation-adjusted revenue per worker rising for the seven and remaining broadly flat for the 493. Profit margins for the seven reached roughly 30%, while the comparison group remained near 10%. The figures do not establish that AI caused the gap: they reflect company-wide performance and do not separate AI from other business or economic factors.

The note also cites San Francisco Federal Reserve productivity measures that, according to Apollo, show no sign of an AI-driven productivity boost so far. Separate Ramp spending data cover its more than 85,000 U.S. business customers, not the S&P 500 as a whole. Ramp’s figures show that the top 10% of firms account for 99.5% of spending on model serving and inference, and 95.6% of spending on OpenAI and Anthropic. Those results suggest AI spending is concentrated, but they do not establish what returns those buyers receive.

Meta’s October 22 filing lists about $355.73 billion in leases not yet started and $454.52 billion in non-cancelable contractual commitments, plus up to $14.72 billion in contingent cloud purchases. The totals include items that are not all AI-related, including some Reality Labs hardware. Meta reported $34.91 billion in capital expenditures in the third quarter and narrowed its 2026 capital spending outlook to $135 billion to $145 billion. Its free cash flow was negative $2.12 billion for the quarter, compared with positive $10.6 billion a year earlier.

Returns Are Hard to See, Costs Are Not

The divide matters because companies are committing substantial sums to computing capacity before broad returns are apparent in economy-wide data. If AI use does not raise revenue or reduce costs enough to cover software, chips and infrastructure, spending could weigh on margins. That is a risk for investors and businesses planning their own deployments, not proof that the spending will fail.

The aggregate picture also has limits. A flat average for 493 firms could conceal companies or departments that benefit, offset by others that do not. It cannot show whether AI caused a particular company’s results. Evidence from specific work processes points to a more mixed outcome: cited customer-support research found issues resolved per hour rose 14% on average and 34% for novice workers, while retail experiments reported sales changes ranging from zero to 16.3%, depending on the workflow. These examples measure particular settings, not broad corporate returns.

Financing adds another pressure point as spending expands beyond established technology firms. The source reports that SpaceX borrowed $40 billion to buy AI chips, while credit protection for Meta and CoreWeave debt became more expensive. Those details signal attention to funding risks, but do not by themselves show that either company cannot meet its obligations or that AI investment is unprofitable.

What Apollo’s Figures Measure

Apollo’s October 24 Daily Spark note, titled “No Signs of AI Payoff Outside the Tech Sector,” drew attention to a gap between the performance of large technology companies and the rest of the S&P 500. The note’s comparisons use company-level revenue per worker and profit margins, alongside productivity indicators. They are descriptive measures; they do not isolate AI adoption or establish a direct link between AI spending and financial results.

The note also draws on Ramp data to show that AI-related spending is concentrated among a small group of companies. Apollo says 80.4% of the top 1% of AI spenders are venture-backed, compared with 13.1% of median spenders. That is a profile of Ramp customers, rather than a survey of every U.S. business. Slok has characterized some venture-funded token spending as “circularity,” but that is his interpretation of how the market is financed.

Meta’s filing provides a separate view of the scale and timing of infrastructure obligations. Its commitments include leases and contracts extending over several years, with some lease arrangements running through 2036. The roughly $825 billion headline total is not an amount Meta must pay all at once, nor is it entirely an AI bill. Forecasts that the five largest technology spenders could invest nearly $780 billion in 2026 and more than $1 trillion annually from 2027 through 2030 are projections, not contractual commitments.

“No Signs of AI Payoff Outside the Tech Sector”

— Torsten Slok, Apollo chief economist, in the October 24 Daily Spark note

What the Aggregate Data Cannot Show

Apollo’s figures do not establish that AI has produced zero returns for every non-tech company. A sector-wide average can mask gains and losses, and the data cited do not separate AI’s contribution from interest rates, labor costs, tariffs or other changes. The note also does not provide a company-by-company assessment of AI projects.

It remains unclear how much of Meta’s reported commitments will be used directly for AI, how quickly the obligations will turn into cash payments, and what revenue or cost savings the company expects them to generate. The cited workplace studies measure individual tasks or trials; their results cannot be treated as a general return across industries. The source’s claim that 83% of nearly 9,000 executives report no productivity impact so far is a survey result, not a measure of productivity across the whole economy.

Credit insurance costs and borrowing figures indicate that investors are watching financing needs, but the available information does not show how the reported costs will change or whether they will constrain future construction. Longer-term returns will depend on adoption, use, pricing and the costs of operating the systems.

Spending and Productivity in Focus

Investors and company leaders will look for evidence in forthcoming earnings reports: whether AI spending is lifting revenue, lowering operating costs or improving output per worker. Meta’s next financial updates should clarify how its capital spending develops against its $135 billion to $145 billion 2026 outlook, although the commitments in its filing span multiple years.

Further productivity data and studies of specific deployments may help distinguish isolated task gains from broad changes in company performance. For now, Apollo’s conclusion is best read as a statement about what is visible in selected aggregate indicators—not a final verdict on AI’s returns across every business.

Key Questions

What did Apollo mean by “no payoff outside tech”?

Apollo said revenue per worker and profit margins were rising for the seven largest technology companies it tracks but broadly flat for the other 493 S&P 500 firms. That comparison describes aggregate trends; it does not prove that every non-tech company has received no benefit from AI.

Does Meta have to pay $825 billion now?

No. The figure combines future lease and contractual commitments, plus contingent cloud purchases. Many extend across several years, and the total is not entirely AI-related or due at once.

Is there evidence AI improves work?

Some task-level studies cited in the source report gains, including more customer-support issues resolved per hour and sales increases in some retail experiments. Results vary by workflow and do not establish economy-wide returns.

Why is AI spending concentrated?

Ramp data cited by Apollo show that a small share of its business customers accounts for most spending in several AI categories. The figures cover Ramp’s customer base, not all U.S. businesses, and do not explain the returns those companies receive.

What should readers watch next?

Future company earnings and productivity data may show whether AI spending is translating into higher revenue, lower costs or measurable output gains. The timing and scale of any broad effect remain uncertain.

Source: ThorstenMeyerAI.com